When people speak of IBM Watson, they do so in hushed tones in the corner of rooms, whispering for fear it will jinx it and the promise will not be realized. ‘Did you hear? Watson will cure cancer.’ ‘Did you hear? IBM Watson can turn water into wine.’ Watson is, in many ways, the Chuck Norris of cognitive computing.
Watson’s reputation stems, in no small part, from IBM itself, who are not coy when it comes to blowing their own trumpet. Dr. John Kelly, who heads up the Watson team at IBM, has claimed that, ‘This is brand new technology. It’s going to change the world.’ Not to be outdone, Ann Rubin, IBM's VP of branded content and global creative, said that Watson enables users to ‘outthink cancer, outthink risk, outthink doubt, outthink competitors.’ Big promises. But is this hype warranted, or, as some believe, is the emperor running amok completely naked?
IBM has long been at the forefront of the AI revolution. It established itself as a leader in the field back in 1996 when its Deep Blue chess playing computer defeated Russian chess Grandmaster Garry Kasparov, while Watson’s Jeopardy victory in 2011 cemented its reputation. Shortly after the Jeopardy victory in 2014, IBM broke down its capabilities into 40 different components that each served a particular business problem, which is the business model it still uses today.
IBM pitches Watson as a solution for a variety of industries, from healthcare to financial services. It crunches through vast amounts of data in order to understand problems and environments in the same way humans do, a process known as cognitive computing. It then uses these insights to provide predictions and make decisions. According to IBM, the business is growing. It is at the center of the company's ‘strategic imperatives’ - which include cloud, mobile and analytics - that contributed $32.8 billion in revenue in 2016. However, further details around ROI are scant, as it is lumped in with the other technologies. In February last year, chief executive Ginni Rometty explained that it does not provide much information because Watson is new and growing, telling the annual IBM analyst meeting, ’We are building an era, a platform, an industry, and making a market with it. We have competitors who don’t disclose for a decade, [so] I’m going to protect it and nurture it — we will disclose eventually’.
Rometty’s position is understandable, but evidence of success may at least help to counter the growing body of criticism. Among these critics is Meg Whitman, CEO of IBM competitor Hewlett-Packard Enterprise. Last year she noted that, ‘We're in a lot of customers where actually from a Watson perspective it's not as far along in terms of real-world applications as you might imagine from the advertising.’ This could easily be dismissed as a competitor trying to get one over on a rival, but her’s is not a lone voice. Many others are starting to argue that Watson is just a thinly veiled advertising campaign, exploiting the notoriety achieved from its Jeopardy! win to promote its consultancy services and the raft of technologies it has put under the Watson umbrella that are not especially groundbreaking.
Also among the dissenters is Oren Etzioni, head of the AI lab set up by Microsoft co-founder Paul Allen, who has argued that, ‘We have no evidence that IBM is able to take that narrow success and replicate it in broader settings. I’m not aware of a single, super-exciting app.’ Founder of the Institute for the Learning Sciences at Northwestern University, Roger Schank Ph.D, even goes so far as to flat out accuses Watson of false advertising, writing in an excoriating blog post that Watson is essentially just a ‘word counter’ little different from Google search. He says that it is incapable of any non-trivial reasoning, contrary to how IBM presents it in its advertising, noting specifically an ad campaign IBM ran that showed Watson’s analysis of Bob Dylan songs for key themes, which it identified as ‘love fades’ and ‘time passes’ - themes Schank argued no Dylan fan would ever put down as his most important. Although, Dylan says those themes sound ‘about right’ in the advert, so it could just be that Schank is misremembering.
Reviews from developers are perhaps more damning for Watson’s longterm success. Watson requires a significant amount of data to confirm and reconfirm its best answer, which very few domains are capable of storing, and it reportedly needs a lot of work by the customers themselves and struggles with real-world, messy data. The Wall Street Journal has revealed that ‘Watson’s basic learning process requires IBM engineers to master the technicalities of a customer’s business and translate those requirements into usable software. The process has been arduous.’ IBM is now facing strong competition in the space and will need to correct such problems quickly. Microsoft Research has spent the last few years incubating a cognitive platform under the name Project Oxford, which it recently made generally available as Microsoft Cognitive Services, now the AI and cognitive computing arm of Microsoft’s Cortana Intelligence suite. Google has also redoubled its efforts in the last few years, developing platforms that include Natural Language Processing, Vision and Speech APIs. Both offerings look very similar to Watson’s in terms of functional capabilities. Although IBM has an advantage because it has been around for longer, the significant resources their rivals can deploy means that IBM will have to work hard to stay ahead of the pack.
IBM refutes accusations that it is not the high end proposition it claims. Dr Kelly says that, ‘Everything we brand Watson analytics is very high-end AI,’ and you will find little argument that IBM’s Watson artificial intelligence system is an incredible piece of technology. It’s more than capable of searching across vast repositories of unstructured digital data and returning answers remarkably quickly, and has done amazing things, particularly in healthcare and refining diagnoses. The problem is not whether its a market leader, though, it’s the promises it is making. Watson has suffered from the goals it has set for itself, from curing to cancer to being the solution to cybersecurity. These are important things to resolve and IBM need to be honest with how much Watson can really do. Overhyping AI is dangerous, as we saw when inflated expectations and subsequent disillusionment in the 1980s led to the so-called AI Winter, bringing investment to a grinding halt and pushing research ‘underground’. Obviously, IBM wants to sell its product, and because it is such a nascent technology it needs big targets like cancer and cybersecurity to put itself on the map and gain customer support. However, in the long term it could be shooting itself in the foot. The thaw has been hard earned, and although another setback of the same magnitude is unlikely for AI, expectations should still be managed.